Forums / Community / Matchmaking Feedback & Discussion

[Locked] MATCHMAKING FEEDBACK UPDATE – August 31

OP ZaedynFel

  1. 1
  2. ...
  3. 9
  4. 10
  5. 11
  6. 12
  7. 13
  8. ...
  9. 26
ZaedynFel wrote:
Did we test K/D and Spread?Yes, they were both worse than separating KPM and DPM. Though having KPM and DPM pretty much gives you spread. Remember, we only look at KPM and DPM within 1 match. We don't track their average across games or anything like that. The influence that the two have is learned on a ton of data, and learned per mode.

Updating just based on relative MMRs?It has to look at CSR as well, and yes that's what it did before. It leads to mass inflation if you can get carried or if you can force an unusually high win %.
Thanks for the response. I guess it affects people differently, but in my personal experience I had around a 50% win rate in the times of inflation (ie every season up until the last) and 68% win percentage with the new roll out. Maybe my experience is the odd one out and not the norm.
@ZaedynFel
Is there a possibility of getting party restrictions in the future? Halo Master Chief Collection currently uses party restrictions and the population of that game is smaller than that of H5. After my matchmaking experience on MCC, I can say it is way better when teams are forced to match teams. When I search alone it's always pretty competitive and fair, and when I search 4s matching those other sweaty teams makes it great. The community thrives at its best on party restrictions for both the people in 4s and the people going in alone or in smaller parties. It'd be cool if next season on H5 if party restrictions was just "tested" at least. Why not just try it out instead of assuming it won't work? It's not like people are gonna quit playing H5 if party restrictions doesn't work out, anyone who's been playing this long is gonna keep playing it no matter what.
OnAllDrugs wrote:
@ZaedynFel
Is there a possibility of getting party restrictions in the future? Halo Master Chief Collection currently uses party restrictions and the population of that game is smaller than that of H5. After my matchmaking experience on MCC, I can say it is way better when teams are forced to match teams. When I search alone it's always pretty competitive and fair, and when I search 4s matching those other sweaty teams makes it great. The community thrives at its best on party restrictions for both the people in 4s and the people going in alone or in smaller parties. It'd be cool if next season on H5 if party restrictions was just "tested" at least. Why not just try it out instead of assuming it won't work? It's not like people are gonna quit playing H5 if party restrictions doesn't work out, anyone who's been playing this long is gonna keep playing it no matter what.
We already have party restrictions. They're built into the skill service. Currently, when to4 play against, e.g., 4 solo players, they are NOT winning more than expected. The skill system now knows near exactly how much advantage the parties have and compensates for it automatically in the matchmaking. This is why it takes forever for high-end parties to find matches --- because you would need another high end party.

But there's no reason to make an average level Plat/Gold team of 4 wait for another to4 when you can get the same fairness vs. 4 solo Diamonds etc.

Also, we did previously test using old school party restrictions and it "didn't work out" as you put it.
@ZaedynFel Is there a way to merge kpm and kda weights for determining mmr? Statistically kpm is better at predicting wins but imo still doesn't represent skill fully. If kda and perhaps medals per minute could be tied in I think it'd be perfect or at least more integrible. Perhaps use kpm as some sort of baseline and refine it with kda and mpm? Im not a programmer so idk if any of that works but I feel like kda and medals should be involved somehow. If someone gets 27 kills but 35 deaths in slayer they've failed their team
This post has been edited by a moderator. Please do not create alternate accounts to bypass forum bans. Alternate accounts will be permanently banned, and offending users will be subject to both temporary and permanent bans.
*Original post. Click at your own discretion.
Spoiler:
Show
Sturbz wrote:
@ZaedynFel Is there a way to merge kpm and kda weights for determining mmr? Statistically kpm is better at predicting wins but imo still doesn't represent skill. If kda and perhaps medals per minute could be tied in I think it'd be perfect or at least more integrible. Perhaps use kpm as some sort of baseline and refine it with kda and mpm?
KDA isn't as predictive, and doesn't add anything given kpm and dpm.

Predicting wins is the definition of skill. What other definition would we use other than your ability to win matches?

Medals per minute isn't predictive beyond what we have (tested it, and tested many medals individually).

Also, it sounds like you're thinking that we calculate skill like "skill = kpm x weight" but it's nothing like that.

The strongest signal in the current model is still "Did you win the match?" and kpm/dpm act as rails against where it's likely you should be.
This post has been edited by a moderator. Please do not create alternate accounts to bypass forum bans. Alternate accounts will be permanently banned, and offending users will be subject to both temporary and permanent bans.
*Original post. Click at your own discretion.
Spoiler:
Show
This post has been edited by a moderator. Please do not create alternate accounts to bypass forum bans. Alternate accounts will be permanently banned, and offending users will be subject to both temporary and permanent bans.
*Original post. Click at your own discretion.
Spoiler:
Show
@ZaedynFel don't you think they should being back griftball?
I don't have the MCC data in front of me to look at. I would need to see how big their population is and how many playlists they already have.

It looks like they have almost as many playlists as we do currently, so they might be waiting to see how pop goes.
ZaedynFel wrote:
Sturbz wrote:
@ZaedynFel Is there a way to merge kpm and kda weights for determining mmr? Statistically kpm is better at predicting wins but imo still doesn't represent skill. If kda and perhaps medals per minute could be tied in I think it'd be perfect or at least more integrible. Perhaps use kpm as some sort of baseline and refine it with kda and mpm?
KDA isn't as predictive, and doesn't add anything given kpm and dpm.

Predicting wins is the definition of skill. What other definition would we use other than your ability to win matches?

Medals per minute isn't predictive beyond what we have (tested it, and tested many medals individually).

Also, it sounds like you're thinking that we calculate skill like "skill = kpm x weight" but it's nothing like that.

The strongest signal in the current model is still "Did you win the match?" and kpm/dpm act as rails against where it's likely you should be.
Fair enough. Thanks again for all your time and responses to our feedback
have you guys tested KAPM? KPM + (APM/3)?
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
vibr8high wrote:
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
Unfortunately winning games doesn’t constitute true skill. Instead you need to have the stats that best predict wins.
bearhound wrote:
vibr8high wrote:
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
Unfortunately winning games doesn’t constitute true skill. Instead you need to have the stats that best predict wins
But that doesnt make since... How can winning not constitute true skill but losing does.. A win means nothing.. but a loss means something??? I win.... bar doesnt move.. I lose bar moves.. What?

Thats like if I have a bank account and I put money in it and my balance doesnt go up but when I take money out it subtracts from my balance... doesnt make sense...
But beating players far less skilled than you to achieve ranks higher than players far more skilled than you without having to face said highly skilled players makes the ranks meaningless. I suppose that screams "True Skill" for some, though.
But beating players far less skilled than you to achieve ranks higher than players far more skilled than you without having to face said highly skilled players makes the ranks meaningless. I suppose that screams "True Skill" for some, though.
This. One of my friends grinded to Champion a couple of seasons back by playing Platinum players for the whole season, because his MMR largely stayed Platinum/Low Diamond. Huge MMR/CSR discrepancies like that is why True Skill 1.0 ranks was never reflective of actual skill, it was simply an indication of how much you played in the season. Now they did mention that it is possible that Halo Infinite will have a "Team Rank" which you can grind for getting wins like under the old system, but this is impossible for Halo 5 since they are focusing their development time on Infinite.

Now this will come out as harsh but the reality is that a lot of the complainers in these threads have had their egos hurt because of their inflated ranks from past seasons when they were not actually as good as the system told them they were. Now I don't blame them for feeling that way, because they've been playing under that system for most of the game's lifespan.

I believe that when Halo Infinite launches with True Skill 2.0 off the bat the complaints will go away for the most part.
But beating players far less skilled than you to achieve ranks higher than players far more skilled than you without having to face said highly skilled players makes the ranks meaningless. I suppose that screams "True Skill" for some, though.
This. One of my friends grinded to Champion a couple of seasons back by playing Platinum players for the whole season, because his MMR largely stayed Platinum/Low Diamond. Huge MMR/CSR discrepancies like that is why True Skill 1.0 ranks was never reflective of actual skill, it was simply an indication of how much you played in the season. Now they did mention that it is possible that Halo Infinite will have a "Team Rank" which you can grind for getting wins like under the old system, but this is impossible for Halo 5 since they are focusing their development time on Infinite.

Now this will come out as harsh but the reality is that a lot of the complainers in these threads have had their egos hurt because of their inflated ranks from past seasons when they were not actually as good as the system told them they were. Now I don't blame them for feeling that way, because they've been playing under that system for most of the game's lifespan.

I believe that when Halo Infinite launches with True Skill 2.0 off the bat the complaints will go away for the most part.
i was onyx with TS1 and i am onyx with TS2, the only problem i have now 80 win ratio and a K/D near 2. I found many champion last season who have played only against platinum, i will tell you more it more easy now than before.
vibr8high wrote:
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
TrueSkill is the label given to the algorithm that determines your MMR which is used by the Matchmaker and Team Balancer to form “fair” skill-based matches. It’s unlikely you have an issue with it per se because this current version (2.0) is providing vastly superior SBMM over its previous iteration.

Competitive Skill Ranks (CSR) is the label given to the in-game visible reflection of your skill. It’s been redesigned to better follow and remain within a closer range of your MMR while maintaining some specific caveats such as winning will always result in upward movement while losing constitutes a downward movement. You seem to have an issue with this because it’s what you’re visually experiencing and you don’t seem to like the design aspect that keeps it in closer range to your MMR.

But beating players far less skilled than you to achieve ranks higher than players far more skilled than you without having to face said highly skilled players makes the ranks meaningless. I suppose that screams "True Skill" for some, though.
This. One of my friends grinded to Champion a couple of seasons back by playing Platinum players for the whole season, because his MMR largely stayed Platinum/Low Diamond. Huge MMR/CSR discrepancies like that is why True Skill 1.0 ranks was never reflective of actual skill, it was simply an indication of how much you played in the season. Now they did mention that it is possible that Halo Infinite will have a "Team Rank" which you can grind for getting wins like under the old system, but this is impossible for Halo 5 since they are focusing their development time on Infinite.

Now this will come out as harsh but the reality is that a lot of the complainers in these threads have had their egos hurt because of their inflated ranks from past seasons when they were not actually as good as the system told them they were. Now I don't blame them for feeling that way, because they've been playing under that system for most of the game's lifespan.

I believe that when Halo Infinite launches with True Skill 2.0 off the bat the complaints will go away for the most part.
The “Team Rank” discussion that’s been had previously related more to the distribution of a team’s net MMR adjustment amongst the team members, not really about the CSR. In other words, instead of distributing the team's net MMR adjustment based on the individual performances of the team members the net adjustment would get divided up equally amongst the team members regardless to their individual performances. This would require a separate CSR that could specifically follow that group of players who would register as a team unit. CSR adjustments though have, more or less, been a separate topic. One important change for CSR adjustments that hopefully will be realized in Infinite is the removal of the 1-match lag that currently exists.
eLantern wrote:
vibr8high wrote:
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
TrueSkill is the label given to the algorithm that determines your MMR which is used by the Matchmaker and Team Balancer to form “fair” skill-based matches. It’s unlikely you have an issue with it per se because this current version (2.0) is providing vastly superior SBMM over its previous iteration.

Competitive Skill Ranks (CSR) is the label given to the in-game visible reflection of your skill. It’s been redesigned to better follow and remain within a closer range of your MMR with some specific caveats such as winning will always result in upward movement while losing constituents downward movement. You seem to have an issue with this because it’s what you’re visually experiencing and you don’t seem to like the design aspect that keeps it in closer range to your MMR.
I think the issue is that some of us are on the cusp to get to that next rank and yet when we make it over and win 10 games in a row in the higher rank, then its erased if you lose 2 games in a row. Now you have told me and others we must improve and it is the KPM average we're just under. Well I'll say it again, it's hard to get better if you hardly get to play more skilled players. That's what is frustrating the most. Then having to claw your way back up again, just to lose it again cause some dip $#it went 4-18. That's what drives us nuts. The rank sytem does not see any of this to make its decisions and purely goes by numbers.
I and all of us understand that the people working on this, are experts. Masters of their craft. Brilliant people. That being said, it does not make it free of issues. This fluctuation between ranks being one.
On a previous thread I mentioned something about when reaching the next tier, to have us do another 10 placement games which a monitor(I believe it was) said that no one would wanna do that. So I'll try and explain it again, given this new info about how TS2 works.
So when a player gets to the next rank, I think TS2 should immediately disregard the games previous and should let you play 10 placement matches in the new tier. It does not have to be visible to the player. It can just say diamond or platinum, etc. Then after that, let it see how you did. Was KPM on par? Did anyone quit during these games that caused a big swing in the wrong direction? I feel like getting into the new tier and then having the game judge someone on that is better than taking all the numbers from way, way back might help. It can maybe even stop smurfs from placing down low, just so they can smash people. Cause if they do the rank system would see that all of the sudden they're going 30-3 for all these games in a row or fluctuating abnormally high or low every other game.
Thoughts anyone? Not for h5 cause we don't want to take away from h6, but if someone could elaborate pros and cons about this idea that aren't just a naysayer to outside ideas that would be awesome 😁
eLantern wrote:
vibr8high wrote:
I dont care what anyone says... I dont care about statistics and all that... If I win 5 games in a row and my bar barely moves and I lose one and it goes down significantly.. that is not a reflection of TRUE SKILL... Change the name of the ranking system because it is not true skill... and also I get ranked with kids who have higher csr but dont do better than me..... regularly... but I can't seem to go up because I lose one game out of 10...TRUE skill my -Yoink-...
TrueSkill is the label given to the algorithm that determines your MMR which is used by the Matchmaker and Team Balancer to form “fair” skill-based matches. It’s unlikely you have an issue with it per se because this current version (2.0) is providing vastly superior SBMM over its previous iteration.

Competitive Skill Ranks (CSR) is the label given to the in-game visible reflection of your skill. It’s been redesigned to better follow and remain within a closer range of your MMR with some specific caveats such as winning will always result in upward movement while losing constituents downward movement. You seem to have an issue with this because it’s what you’re visually experiencing and you don’t seem to like the design aspect that keeps it in closer range to your MMR.
I think the issue is that some of us are on the cusp to get to that next rank and yet when we make it over and win 10 games in a row in the higher rank, then its erased if you lose 2 games in a row. Now you have told me and others we must improve and it is the KPM average we're just under. Well I'll say it again, it's hard to get better if you hardly get to play more skilled players. That's what is frustrating the most. Then having to claw your way back up again, just to lose it again cause some dip $#it went 4-18. That's what drives us nuts. The rank sytem does not see any of this to make its decisions and purely goes by numbers.
I and all of us understand that the people working on this, are experts. Masters of their craft. Brilliant people. That being said, it does not make it free of issues. This fluctuation between ranks being one.
On a previous thread I mentioned something about when reaching the next tier, to have us do another 10 placement games which a monitor(I believe it was) said that no one would wanna do that. So I'll try and explain it again, given this new info about how TS2 works.
So when a player gets to the next rank, I think TS2 should immediately disregard the games previous and should let you play 10 placement matches in the new tier. It does not have to be visible to the player. It can just say diamond or platinum, etc. Then after that, let it see how you did. Was KPM on par? Did anyone quit during these games that caused a big swing in the wrong direction? I feel like getting into the new tier and then having the game judge someone on that is better than taking all the numbers from way, way back might help. It can maybe even stop smurfs from placing down low, just so they can smash people. Cause if they do the rank system would see that all of the sudden they're going 30-3 for all these games in a row or fluctuating abnormally high or low every other game.
Thoughts anyone? Not for h5 cause we don't want to take away from h6, but if someone could elaborate pros and cons about this idea that aren't just a naysayer to outside ideas that would be awesome 😁
It doesn't matter who you play, you just have to get a KPM that is better than your current MMR against any lobby to get your MMR to go up. Whether you are facing Platinum, Diamond, Onyx etc. does not matter. Your suggestion of facing people that are significantly better than you (IE. Those that you won't be matching) will not make you a better player. You will just get destroyed in blowout games and get frustrated. The way to get better is to play against progressively better people (small increases in skill gap) and that will happen in the current system if you start doing a bit better each game. If you perform to par with what the True Skill thinks then that means it is absolutely correct about your rank.
  1. 1
  2. ...
  3. 9
  4. 10
  5. 11
  6. 12
  7. 13
  8. ...
  9. 26